Shape-based image retrieval of Chinese paper-cutting using RBFNN with invariant moment

Computer aided design (CAD) system for traditional Chinese paper-cutting provides significant assistance for folk artists creating delicate handicrafts. However, the large amount of paper-cutting patterns creates several major challenges for the management of an online pattern library. Artists could easily collect favorite pattern materials and share their creative works with others using an online pattern library. A proper image descriptor and effective content-based image retrieval (CBIR) strategy are needed to provide a resourceful online database. We propose to combine invariant moments descriptor of paper-cutting patterns and Radial Basis Function Neural Networks (RBFNN) trained by a minimization of the Localized Generalization Error (LGEM) to provide the CBIR function for paper-cutting retrieval. Experimental results show that the proposed method outperforms similarity based method.